Integrations
PicoClaw
Use Router9 as an LLM provider in PicoClaw
PicoClaw is an ultra-efficient AI assistant written in Go with multi-channel support. You can add Router9 as a custom provider via the model list configuration.
Setup
Edit your PicoClaw config at ~/.picoclaw/config.json and add Router9 to the model_list:
{
"version": 2,
"model_list": [
{
"model_name": "router9-gpt4o",
"model": "openai/gpt-4o",
"api_base": "https://api.router9.com/v1",
"api_key": "sk-r9k-your-key-here"
},
{
"model_name": "router9-sonnet",
"model": "openai/claude-sonnet-4-20250514",
"api_base": "https://api.router9.com/v1",
"api_key": "sk-r9k-your-key-here"
}
],
"agents": {
"defaults": {
"model_name": "router9-gpt4o"
}
}
}Using Environment Variables
Instead of hardcoding the key, use environment variables:
export PICOCLAW_AGENTS_DEFAULTS_MODEL_NAME=router9-gpt4oOr set PICOCLAW_CONFIG to point to a separate config file:
PICOCLAW_CONFIG=/etc/picoclaw/production.json picoclaw gatewayMultiple Models
Add as many Router9-backed models as you need — they all share the same API key and base URL:
{
"model_list": [
{
"model_name": "fast",
"model": "openai/gpt-4o-mini",
"api_base": "https://api.router9.com/v1",
"api_key": "sk-r9k-your-key-here"
},
{
"model_name": "smart",
"model": "openai/gpt-4o",
"api_base": "https://api.router9.com/v1",
"api_key": "sk-r9k-your-key-here"
},
{
"model_name": "reasoning",
"model": "openai/claude-sonnet-4-20250514",
"api_base": "https://api.router9.com/v1",
"api_key": "sk-r9k-your-key-here"
}
]
}Load Balancing
PicoClaw supports round-robin load balancing. You can use multiple Router9 agent plan keys for the same model:
{
"model_list": [
{
"model_name": "gpt4o",
"model": "openai/gpt-4o",
"api_base": "https://api.router9.com/v1",
"api_key": "sk-r9k-key-agent-1"
},
{
"model_name": "gpt4o",
"model": "openai/gpt-4o",
"api_base": "https://api.router9.com/v1",
"api_key": "sk-r9k-key-agent-2"
}
]
}Tips
- Use the
openai/vendor prefix since Router9 exposes an OpenAI-compatible API. - Store your API key in PicoClaw's
.security.ymlfile instead of the main config to avoid accidental exposure. - See the PicoClaw model configuration docs for advanced options like timeouts and rate limits.